This happens in Windows, running darcs 1.0.8 in cmd.exe, but not when running
the same darcs binary from a cygwin shell, or in linux.

The file longline.txt is a 120KB text file with no line breaks.
C:\longline>darcs init
C:\longline>darcs add longline.txt
C:\longline>darcs record -A me
addfile ./longline.txt
Shall I record this change? (1/?)  [ynWsfqadjkc], or ? for help:
hunk ./longline.txt 1
-
+
darcs failed:  resource exhausted
C:\longline>

I guess the problem is that we're writing the very long line to the terminal,
and that causes trouble.  Perhaps (for windows only?) we should truncate lines
when writing to the terminal?

Can you tell me if

record -A me -a

works fine?

Yes, record -a works fine.

Edwin,

Could you test with one of the Darcs2 pre-release binaries for Windows and see
if the experience is improved? They are available here:

http://wiki.darcs.net/index.html/DarcsTwo

  Mark

I'd guess this problem is still there.  I think we need to find out what limit
cmd.exe places on length of lines printed.  Someone with windows ought to be
able to run a simple test.  Of course, it may depend on the variety of windows run.

David

I found these references about long lines and cmd.exe, but they are both mainly
about how long the command line call can be, and how long the environment
variables can be (which I didn't think we used on Windows).

http://blogs.msdn.com/oldnewthing/archive/2003/12/10/56028.aspx
http://support.microsoft.com/kb/830473

We could find out the dumb way with a little program to print out increasingly
long lines to screen...

Edwin: do you think this Haskell program could give us any clues? Could you, or
somebody else on Windows tell us what happens when you run it?

Well, I just got access to a Windows box... and the longlines.hs program
suceeeds.  Hmm...

We still need to diagnose this bug, after my longlines.hs failed to confirm the
writing-long-lines-to-terminals hypothesis from David (of course, it could just
be that the test itself needs to be tweaked)